CN109583095B

CN109583095B - North Pacific typhoon extension period forecasting method based on hybrid statistical power model

Info

Publication number: CN109583095B
Application number: CN201811464661.4A
Authority: CN
Inventors: 徐邦琪; 钱伊恬
Original assignee: Nanjing University of Information Science and Technology
Current assignee: Nanjing University of Information Science and Technology
Priority date: 2018-11-30
Filing date: 2018-11-30
Publication date: 2023-04-28
Anticipated expiration: 2038-11-30
Also published as: CN109583095A

Abstract

The invention relates to a North Pacific typhoon extension period forecasting method based on a hybrid statistical power model, which utilizes a modern statistical method to classify the track of historical typhoons, and establishes a statistical forecasting equation of intra-season oscillation signals and typhoons by the influence of intra-season scale sea gas states on different types of typhoons; and forecasting typhoons generated in 10-30 days in the future by using a high-resolution sea-air coupling power mode as a forecasting factor for the forecasting field of the oscillating signals in the seasons of 10-30 days in the future. And finally, multiplying the generation number of typhoons of different types by the climate probability distribution of the track of the typhoons, so that the probability distribution map of the generation and the frequency of the typhoons on the entire North Pacific ocean is predicted 10-30 days in advance. The beneficial effects are that: the total number of typhoons on the North and North Pacific ocean can be effectively predicted in the extension period of the typhoons, which is 10-30 days in the future, and the spatial probability distribution of typhoons generated/frequency in 10 days can be obtained by utilizing different generation positions and movement tracks of the typhoons.

Description

North Pacific typhoon extension period forecasting method based on hybrid statistical power model

Technical Field

The invention relates to the technical field of atmospheric science, in particular to a method for forecasting the extension period of North Pacific typhoon based on a hybrid statistical power model.

Background

Typhoons cause typhoons, storm, billows, storm surge, landslide and mud-rock flow caused indirectly and other geological disasters, and often cause serious casualties and socioeconomic losses. North Pacific ocean is the sea area with the highest typhoon generation frequency, and accounts for about 30% of tropical cyclone numbers in the global sea area. Typhoons develop and move to the west/northwest after being generated in the western pacific of the tropical ocean, and possibly attack southeast coastal areas of China, and statistics show that 6-7 typhoons land on the coastal provinces of China on average in 1983-2006 each year, resulting in economic losses of hundreds of billions per year and casualties of thousands of people per year. Therefore, research and development of typhoons monitoring and forecasting key technologies become important demands for national disaster prevention and reduction, socioeconomic policy formulation and the like.

The typhoon forecasting system of each large business forecasting unit in China and the world at present comprises: 1) Medium-short term (within 5 days) high resolution numerical mode prediction and 2) climate prediction on the rose scale. The typhoon forecasting accuracy for more than one week is lower due to noise of the atmosphere and systematic errors of the numerical mode (Vitart et al 2010); the global climate mode resolution for carrying out typhoon season prediction is low, the mesoscale dynamic process of typhoon cannot be analyzed, and the strength and path simulation capability of typhoon are insufficient. Thus, in addition to the numerical model, rose predictions for typhoons are typically statistical prediction models (Gray 1984;Chan et al.1998;Fan and Wang 2009) and hybrid-statistical models (Murakami et al 2016) built based on statistical relationships of large scale sea states to typhoons generation.

The short-term weather forecast and the weather forecast have obvious gaps, and the weather forecast mode in the extension period (10-30 days) is developed and improved, so that a seamless forecast system is completed, and the method is a primary task of the current global weather and weather forecast research (Waliser 2005). The predictability of the extension forecast comes mainly from intra-season oscillations in the atmosphere (Madden and Julian 1994; li Chongyin et al 2003;Waliser 2005), the intra-season oscillations activity of the tropical-subtropical zone has a significant impact on north-west pacific typhoons (Ding Yihui et al 1977;Gray 1979;Liebmann et al.1994;Maloney and Hartmann 2000; zhou Congwen et al 2004;Kim et al.2008; sun Chang et al 2009; li Chongyin et al 2012; he Jielin et al 2013), and when the intra-season oscillations are in convection phase, both the low frequency cyclonic circulation and the irradiance zone contribute to the weather scale disturbances to gain kinetic energy from the intra-season oscillations, thus more typhoons occur and are enhanced (Chen Guanghua and Huang Ronghui 2009;Hsu et al.2011). Camargo et al (2009) and Zhao et al (2015) also indicate that the mid-level water vapor field and the low-level vorticity field of intra-season dimensions are closely related to low frequency changes in typhoon activity, and thus, it is possible to forecast the extension period dimensions of typhoons based on the correlation of intra-season oscillations and typhoon generation.

Although past studies have found the importance of intra-season oscillatory activity to typhoon activity occurrence and development, methods for applying the correlation of both to the northwest pacific typhoon activity extension forecast have not been established, and current literature only has statistical forecast (Leroy and Wheeler 2008) and power pattern assessment (Vitart et al 2010) of weekly changes in tropical cyclone in the southern hemisphere, and numerical pattern studies of few typhoons in indian, atlantic and northwest pacific, the results of which show that if the pattern can simulate the correct intra-season oscillatory signal, typhoons are likely to be forecasted 2-4 weeks in advance (Fu and Hsu 2011;Wu and Duan 2015;Xiang et al.2015). Therefore, the development of the North Pacific ocean extension period forecasting method and model has high application value of meteorological service.

Disclosure of Invention

The invention aims to overcome the defects of the prior art, and provides a North Pacific typhoon extension period forecasting method based on a hybrid statistical power model, which is realized by the following technical scheme:

the method for forecasting the extension period of the North Pacific typhoon based on the hybrid power model is based on a typhoon optimal path data set provided by a combined typhoon early warning center JTWC under the ocean center of the United states navy Pacific meteorological, and comprises the following steps:

step 1), firstly, dividing tropical cyclone data of a JTWC tropical cyclone data set into a plurality of classes according to generation positions and development tracks by using a c-means fuzzy cluster analysis method;

step 2), searching a statistical relation between a low-frequency large-scale field and the number of generated typhoons of various types, and establishing a statistical prediction equation for predicting the number of generated typhoons of each type;

step 3) leading the corresponding forecasting factors in the GFDL mode output large-scale low-frequency field into an empirical history statistics forecasting equation to obtain the number of distance levels in each ten days of typhoons in the forecasting period 2003-2012; adding the average number of historical climates in typhoon seasons to the average number of typhoons distance to obtain the total number of the forecasted typhoons;

step 4) multiplying the total number of typhoons in each type by the average generation position and track distribution probability distribution of the weather in each ten days of typhoons in each type to obtain the generation position and track occurrence probability of typhoons in each ten days respectively, and adding the generation position and track occurrence probability of all typhoons to obtain the generation position and track frequency probability distribution map of the entire North-west Pacific typhoons in each ten days;

in the step 2), in order to establish a forecasting equation with stable forecasting performance, four methods are adopted to define forecasting factors, which are respectively:

the first method for defining the predictor is as follows: carrying out regional averaging in the square boxes of the regions where the historical typhoons are generated most intensively in each forecasting field, wherein the square boxes of the regions are fixed for each type of typhoons;

the second method for defining the predictor: searching a maximum significant positive correlation area frame and a maximum significant negative correlation area frame on the correlation diagrams of the typhoon distance flat number and the low-frequency large-scale fields respectively, and carrying out area averaging respectively, wherein the size and the position of the corresponding area frame are changed for each large-scale environmental field of each typhoon;

third method for defining predictor: in the largest historical area of typhoon generation, calculating the average of positive and negative grid points of the average typhoon distance number and the correlation coefficient of the low-frequency sea-air field passing through the 95% significance test;

fourth method for defining predictor: searching a large-range region passing through the 95% saliency test on the correlation diagram of the typhoon distance flat number and the low-frequency large-scale field, and subtracting the two regions into a forecasting factor if positive and negative correlation regions meeting the conditions exist at the same time;

defining forecasting factors through the four methods in the step 2) to obtain forecasting factors of each type of typhoons, and then establishing an empirical historical statistics forecasting equation by utilizing a multiple linear stepwise regression method for each type of typhoons forecasting factors and the corresponding historical typhoons distance parallel number; and selecting forecasting factors which are optimal in forecasting efficiency and independent of each other by the stepwise regression method, so as to avoid overfitting.

The method for forecasting the extension period of the North Pacific typhoon based on the hybrid statistical power model is further designed in that the coefficient of the positive correlation lattice point in the third method for defining the forecasting factors is 1, and the coefficient of the negative correlation lattice point is-1.

The further design of the method for forecasting the extension period of the North Pacific typhoon based on the hybrid statistical power model is that the generation position and the generation probability of the track of each type of typhoon in the step 4) in each ten day are as shown in the formula (1):

wherein ,

represent C _k Typhoonlike climate probability generated or passed in a 5 ° x 5 ° box of longitude and latitude grid (i, j) within the first ten days of climate averaging, +.>

Represent C _k Typhoons-like frequency generated or passed in the 5 degree by 5 degree square of the longitude and latitude grid (i, j) within the first ten days of weather averaging, +.>

Represent C _k Typhoons were generated or passed over the entire north pacific region within the first ten days of climate averaging.

The further design of the North Pacific typhoon extension period forecasting method based on the hybrid statistical power model is that the generation position and the track occurrence probability of each type of typhoon to be forecasted in each ten days are as shown in the formula (2):

wherein ,

c representing forecast _k Probability of typhoons being generated or passing in a 5 ° x 5 ° box of the longitude and latitude grid (i, j) within the first ten days of the p-th year; />

Represent C _k The number of typhoons generated in the first ten days of the p-th year, N _{Fcsttotal,l,p} Representing the total typhoons generated throughout the north west pacific in the first ten days of p-th; />

Is C in formula (1) _k Typhoon-like climate average probability.

The further design of the North Pacific typhoon extension period forecasting method based on the hybrid statistical power model is that the total probability of typhoons of each type is added to obtain the total probability of typhoons generated or passed on the longitude and latitude grid (i, j) as shown in the formula (3)

Where nk is the total number of clusters of typhoons c-means, and for typhoons in the pacific northwest, nk is 7.

The further design of the North Pacific typhoon extension period forecasting method based on the hybrid statistical power model is that the step 1) further comprises the preprocessing operation of the mode return data: setting forecasting objects as typhoons generation number, generation position and track probability distribution map in each ten days on the North Pacific ocean; setting a modeling object of the method to be 5 months, 16 days and 12 months and 5 days of 1979-2002, wherein the modeling time and the time date of each year correspond to the forecast time and the time date; and forecasting the typhoon number pitch level in each ten days after the seasonal variation of the climate average is removed, wherein the climate average typhoon seasonal circulation component corresponding to the seasonal variation is obtained by averaging in each ten days of 24 years of 1979-2002 of the modeling period.

The method for forecasting the extension period of the North Pacific typhoon based on the hybrid statistical power model is further designed in that the preprocessing operation further comprises the following steps: the method for extracting the intra-season low-frequency components of the large-scale field of the mode return data specifically comprises the following steps of:

step A), removing annual cycle and first three harmonics from a daily large-scale field to obtain a new field;

step B), subtracting the sliding average value of the previous 120 days from the field obtained in the step A) to obtain a distance flat field without internationally changing;

and C) averaging the distance flat field for 10 days according to the date of the forecast time to obtain a low-frequency large-scale environment field corresponding to the typhoon distance flat ten-day data.

The invention has the following advantages:

the prediction results of the four methods for searching for potential key predictors of the present invention are all very close, see fig. 5, illustrating the stability of the method. The result of 0 days in advance (Lead 0) is obtained by taking the observation data of the forecasting period into a forecasting equation, and the forecasting result can be regarded as the forecasting upper limit of the forecasting mode. The prediction result at 10 days in advance is very close to the upper prediction limit of the mode (namely the result of Lead 0), the correlation coefficient reaches 0.45-0.46, the prediction skill starts to decline along with the increase of the prediction advance time, and the result of 20 days in advance shows that the time correlation coefficient skill is 0.21, and the 95% confidence test is passed.

As can be seen from fig. 5, the forecasting results of the four methods for selecting the forecasting factors are similar, and fig. 6 shows the forecasting results of simply performing the set averaging on the four methods. The time sequence of typhoon pitch flat predicted in advance of 0 days is displayed (part a in fig. 6), the correlation coefficient skill of the upper limit of the mode prediction is 0.52, the prediction skill of the mode is still as high as 0.46 (part b in fig. 6) when 10 days are predicted in advance, the correlation coefficient is continuously reduced, the root mean square error is continuously increased along with the advance of the prediction time, the 95% significance test is the standard with the prediction skill, and the effective prediction capability of the northwest pacific typhoon prediction method based on the hybrid statistical power model on the typhoon pitch flat number is 15-20 days. If the total typhoons are obtained by adding the number of the forecasted typhoons to the average seasonal circulation component of the climate, referring to fig. 7, the forecast still shows a high forecast skill 30 days in advance.

Drawings

FIG. 1 is a flow chart of a power-statistics forecasting model for the extended period of North Pacific typhoon.

FIG. 2 is a graph of seven typhoons and an average trajectory in a JTWC typhoons dataset obtained by using a c-means fuzzy cluster analysis method in 1979-2002 month 6-11.

FIG. 3 is a schematic diagram of a power spectrum analysis of seven classes of typhoons from ten days to ten days of the number of planes in a 1979-2002 6-11 month JTWC typhoons dataset.

FIG. 4 is a graph of time correlation coefficients of 1979-2002 low frequency OLR and VWS from flat field and C1 typhoon ten-day from flat number and a schematic diagram of four regions for selecting key predictors.

Fig. 5 is a schematic diagram of the evaluation of the forecasting results of four types of methods for selecting key forecasting factors. Wherein, the a-d part of FIG. 5 is a schematic diagram of time correlation coefficients of the forecasting results of four types of methods for selecting key forecasting factors; the e-h part of FIG. 5 is a root mean square error diagram of the prediction results of four types of methods for selecting key predictors; the i-l portion of FIG. 5 is a graph showing the AUC index of the forecast results for four classes of methods for selecting key predictors.

Fig. 6 is a graph of typhoon distance flat data for the sum of TCall and C1-C7 for mixed forecasts of 5 months from 2003-2012, 16 days-12 months 5 days in advance (part a of fig. 6), 0 days, (part b of fig. 6), 10 days, (part C of fig. 6), 15 days, (part d of fig. 6), 20 days, (part e of fig. 6), 25 days and (part f of fig. 6) 30 days forecast time.

Fig. 7 is a graph of the root mean square error (root mean square error) and probability forecast AUC index (root mean square error of fig. 7, fig. 7) for the average number of typhoons (parts a-c of fig. 7) and the total number (parts d-f of fig. 7) of the mixture forecast and observation (parts a of fig. 7, d of fig. 7).

FIG. 8 is a graph showing the prediction results of four typhoon generation cases (part a-d of FIG. 8), the probability of "perfect reconstruction" of (part e-h of FIG. 8), 0 days in advance (part i-l of FIG. 8), 10 days in advance (part m-p of FIG. 8) and 15 days in advance (part q-t of FIG. 8).

FIG. 9 shows the observation probability (part a-d of FIG. 9), the probability of "perfect reconstruction" of (part e-h of FIG. 9), the prediction results of 0 days in advance (part i-l of FIG. 9), 10 days in advance (part m-p of FIG. 9) and 15 days in advance (part q-t of FIG. 9) for four instances of typhoons.

Detailed Description

The present invention will be described in detail with reference to the accompanying drawings.

Referring to fig. 1, the method for forecasting the extension period of the pacific typhoon in northwest based on the hybrid power model provided by the embodiment is based on a typhoon optimal path data set provided by a united typhoon early warning center JTWC under the united states navy pacific weather and ocean center, and comprises the following steps:

step 1) firstly, classifying the JTWC tropical cyclone data set 1979-2002 tropical cyclone data of 5-11 months into seven categories according to generation positions and development tracks by using a c-means fuzzy cluster analysis method, wherein the classification result is shown in figure 2. Each class of tropical cyclone has unique generation positions and development tracks, and the number of typhoons in each class has significant 10-90 days seasonal variation characteristics, see fig. 3.

Typhoon best path dataset: is provided by united typhoon pre-warning centers (Joint Typhoon Warning Center, JTWC) subordinate to the united states navy pacific weather and ocean center. The data time period was from 1979 to 2012 with a 6 hour time interval. According to the tropical cyclone grade definition of Saffir-Simpson, the forecasting object of the method is a tropical storm with the maximum surface wind speed of more than 34 knots.

step 3) leading the corresponding forecasting factors in the GFDL mode output large-scale low-frequency field into an empirical history statistics forecasting equation to obtain the number of distance levels in each ten days of typhoons in the forecasting period 2003-2012; and adding the average number of the historical climates of the typhoon seasonal variation to the average number of the typhoons to obtain the predicted total typhoons.

The sea-air coupling mode with high resolution of GFDL in the United states has 32 layers of vertical layers in the atmosphere mode, 50km in the horizontal resolution, 50 layers of vertical layers in the ocean mode and 1 degree by 1 degree in the horizontal resolution. The climate pattern outputs a prediction result once from

month

1, 6, 11, 16, 21 and 26 of 4-11 of 2003-2012, and takes 00, 04, 08, 12 and 16 of world time per day as different initial conditions, and there are 5 prediction sets eachIntegrate 50 days after. There were 2400 rewards data (10 years x 8 months x 6 days x 5 sets) in total for these 10 years. The method for forecasting the large-scale sea-air field by using the mode comprises the following steps: external long wave radiation (Outgoing Longwave Radiation, OLR), surface temperature (T _s ) 700hPa specific humidity, 500hPa vertical velocity, vertical wind shear, 850hPa divergence and vorticity fields, for a total of 7 thermal and dynamic large scale environmental fields associated with typhoon generation.

And 4) multiplying the total number of typhoons in each type by the probability distribution of the average generation position and track distribution of the weather in each ten days of typhoons in each type to obtain the generation position and track occurrence probability of typhoons in each ten days respectively, and adding the generation position and track occurrence probability of all typhoons to obtain the probability distribution of the generation position and track frequency of the typhoons on the entire North-west Pacific ocean in each ten days.

In the step 2), in order to establish a prediction equation with relatively stable prediction performance, taking a correlation coefficient diagram between the flat number of the C1 typhoons and the OLR and the vertical wind shear as an example, referring to fig. 4, four methods for defining the predictor used in the embodiment are described:

the first method for defining the predictor is as follows: the regions where the historical typhoons generated most intensively in each forecasting field (the box on the part a of fig. 4 and the box on the part e of fig. 4) are subjected to region averaging. The box of the area is thus fixed for each type of typhoon, i.e. the same box is used for different large scale fields of each type of typhoon, each type of typhoon having 7 predictors.

The second method for defining the predictor: a maximum significant positive correlation region box and a maximum significant negative correlation region box are found on the correlation diagrams of the typhoon pitch flat number and the low frequency large scale field, respectively, and region averaging is performed (as part b of fig. 4 and part f of fig. 4, respectively). The size and location of the boxes will vary for each large scale environmental field of each type of typhoon. There are 14 predictors for each type of typhoon.

Third method for defining predictor: and searching for the average positive and negative grid points of the typhoon distance level number and the correlation coefficient of the low-frequency sea-air field passing the 95% saliency test in the history maximum area (determined by the longitude and latitude of the most boundary of the history typhoon generating position, such as part c of fig. 4 and part g of fig. 4), wherein the coefficient of the positive correlation grid point is 1, and the coefficient of the negative correlation grid point is-1. In the method, each typhoon has 7 predictors.

Fourth method for defining predictor: a large range of regions (e.g., part d of fig. 4 and part h of fig. 4) that pass the 95% saliency test are found on the correlation map of typhoon pitch flat number and low frequency large scale field. If positive and negative correlation areas meeting the conditions exist at the same time, the positive and negative correlation areas are subtracted to form a forecasting factor. Thus, in the method, 7 predictors exist for each typhoon type.

Defining forecasting factors through the four methods in the step 2) to obtain forecasting factors of each type of typhoons, and then establishing an empirical historical statistics forecasting equation by utilizing a multiple linear stepwise regression method for each type of typhoons forecasting factors and the corresponding historical typhoons distance parallel number; and selecting the forecasting factors which have the best forecasting efficiency and are mutually independent through the stepwise regression method, so as to avoid overfitting.

The probability distribution of the climatically North Pacific typhoon in the above step 4) is as shown in formula (1):

wherein ,

Represent C _k Typhoonlike was found throughout the North Pacific ocean in the first ten days of weather averagingThe total frequency of the zones generated or passed.

Forecast C in step 4) above _k The probability distribution of typhoons is as shown in formula (2):

wherein ,

Is C in formula (1) _k Typhoon-like climate average probability.

Adding the probabilities of typhoons of each type to obtain the total probability of generation or passing of the typhoons on the entire North Pacific ocean as shown in the specification (3)

As shown in fig. 8, the prediction results of each case are generated for four typhoons, the spatial correlation coefficients of the prediction results 10 days in advance are all greater than 0.5, and the main areas of typhoons generation can be predicted. Where the "perfect reconstruction" probability of the e-h part of fig. 8 represents the typhoon probability distribution obtained by substituting the number of observed typhoons into equation (2). The upper right corner is the spatial correlation coefficient of each predictor with the observed field. As shown in fig. 9, as the forecasting results of four typhoons (tracks), the main track of typhoons moving in 10 days is well forecasted in all the four examples, and the spatial correlation coefficient reaches 0.6-0.8 in 10 days in advance. Where the "perfect reconstruction" probability of part e-h) of fig. 9 represents the typhoon probability distribution obtained by substituting the number of observed typhoons into equation (2). The upper right corner is the spatial correlation coefficient of each predictor with the observed field. The result well illustrates that the method of the invention can not only effectively forecast the extension period of typhoons, but also obtain the spatial probability distribution of typhoons generated/frequency within 10 days by using different generation positions and movement tracks of the seven typhoons.

In the step 1), preprocessing operation is carried out on the return data of the mode: setting forecasting objects as typhoons generation number, generation position and track probability distribution map in each ten days on the North Pacific ocean; setting a modeling object of the method to be 5 months, 16 days and 12 months and 5 days in 1979-2002, wherein modeling time and time of each year correspond to forecast time and time date; and forecasting the typhoon number pitch level in each ten days after the seasonal variation of the climate average is removed, wherein the climate average typhoon seasonal circulation component corresponding to the seasonal variation is obtained by averaging in each ten days of 24 years of 1979-2002 of the modeling period.

The preprocessing operation further comprises: the method for extracting the intra-season low-frequency components of the large-scale field of the mode return data specifically comprises the following steps of:

step A) removing the annual cycle and the first three harmonics from the daily large-scale field to obtain a new field.

Step B) subtracting the sliding average value of the previous 120 days from the field obtained in step A) to obtain a distance flat field without internationally changing.

In the embodiment, after the tracks of the historical typhoons are classified by using a modern statistical method, a statistical prediction equation of intra-season oscillation signals and typhoons is established by the influence of the intra-season scale sea gas state on typhoons of different types. And the dynamic mode is utilized to take a forecasting field of the oscillation signal in the season of 10-30 days in the future as a forecasting factor, and after the forecasting field is substituted into a statistical equation, typhoons generated in 10-30 days in the future can be forecasted. And finally, multiplying the generation number of typhoons of different types by the climate probability distribution of the track of the typhoons, so that the probability distribution map of the generation and the frequency of the typhoons on the entire North Pacific ocean is predicted 10-30 days in advance.

In actual forecasting, the steps 1) and 2) only need to calculate once by using historical observation data, so as to obtain the generation and frequency probability diagrams of typhoons with different track types and the number generated in the history. And utilizing the correlation relation between the generation number of various typhoons in the historical data and the low-frequency large-scale sea state analysis, thereby establishing an empirical forecasting equation of each typhoon type. In the real-time forecasting operation, the low-frequency large-scale forecasting factors of 10-30 days in the future, which are forecasted by the power mode, are only required to be brought into an empirical forecasting equation, the number of distance planes generated by various typhoons in 10-30 days in the future can be obtained, and the probability distribution map of the occurrence of the typhoons in the North Pacific ocean in 10-30 days in the future can be obtained by multiplying the number of distance planes with the weather average probability distribution map of the typhoons in various types and then adding the weather average probability distribution map.

The mixed statistical power model adopted by the method adopts a non-band-pass filtering method to extract the components of the intra-season scale (10-90 days) in the process of establishing, so that the method can be directly applied to real-time prediction. The result of the hybrid statistical power model is the number of typhoons of each type and total typhoons generated on the North-west Pacific ocean in each ten days of the future 10-30 days, and the generation of typhoons and the spatial distribution of the frequency probability on the North-west Pacific ocean.

The present invention is not limited to the above-mentioned embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present invention are intended to be included in the scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for forecasting the extension period of North Pacific typhoon based on a hybrid statistical power model is based on a typhoon optimal path data set provided by a combined typhoon early warning center JTWC under the ocean center of the United states navy Pacific meteorological, and is characterized by comprising the following steps:

2. The method for forecasting the extension period of the pacific typhoon in the northwest based on the hybrid power model according to claim 1, wherein the third method for defining the forecasting factors is characterized in that the coefficient of the positive correlation lattice point is 1, and the coefficient of the negative correlation lattice point is-1.

3. The method for forecasting the extension period of the pacific typhoons in northwest based on the hybrid power model according to claim 1, wherein the occurrence probability of the generation position and the track of each type of typhoons in step 4) in each ten days is as shown in the following formula (1):

wherein ,

4. A hybrid-model-based method for forecasting the extension period of pacific typhoons in the northwest according to claim 3, wherein the generation position and the occurrence probability of the track of each type of typhoons forecasted in each ten days are as shown in the formula (2):

wherein ,

Is C in formula (1) _k Typhoon-like climate average probability.

5. The method for predicting the period of extension of North Pacific typhoons based on a hybrid model according to claim 4, wherein the total probability of typhoons of each type is obtained by summing the probabilities of typhoons of each type, and the total probability of typhoons generated or passing by a longitude and latitude grid (i, j) is shown as formula (3)

6. The hybrid-model-based method for predicting the period of extension of pacific typhoons in northwest of claim 4, further comprising the step of preprocessing the pattern return data in step 1): setting forecasting objects as typhoons generation number, generation position and track probability distribution map in each ten days on the North Pacific ocean; setting a modeling object of the method to be 5 months, 16 days and 12 months and 5 days of 1979-2002, wherein the modeling time and the time date of each year correspond to the forecast time and the time date; and forecasting the typhoon number pitch level in each ten days after the seasonal variation of the climate average is removed, wherein the climate average typhoon seasonal circulation component corresponding to the seasonal variation is obtained by averaging in each ten days of 24 years of 1979-2002 of the modeling period.

7. The hybrid-statistical-power-model-based method for predicting the period of extension of the pacific typhoon in northwest of claim 6, wherein the preprocessing operation further comprises: the method for extracting the intra-season low-frequency components of the large-scale field of the mode return data specifically comprises the following steps of: